Adaptive stimulation design for the treatment of epilepsy

脑刺激最近成为治疗难治性癫痫的一种很有前景的治疗方法。然而，我们对使用电刺激的最佳刺激模式知之甚少，这样我们就能最大限度地减少癫痫发作，同时最大限度地减少对大脑的长期损害。该项目的目标是使用强化学习方法自动优化控制深部脑刺激的闭环策略。强化学习是计算机科学中一个正在进行的研究领域，它可以用于直接从电生理记录中学习和适应刺激策略，并在不同的时间尺度上（从启动对癫痫发作前大脑活动的快速反应，到适应癫痫疾病的长期变化）。我们将使用体外和体内边缘癫痫模型来测试这项新技术。从长远来看，我们预测这项技术将在治疗癫痫患者，特别是内侧颞叶患者方面发挥关键作用。此外，我们有信心，强化学习可以整合到广泛的植入式医疗刺激器中，比如目前用于治疗帕金森症、心脏病和精神病性抑郁症的那些。

Brain stimulation has recently emerged as a promising therapy for patients with medically-intractable epilepsy. However little is known about the best stimulation patterns to use when applying electrical stimulation, such that we get maximal seizure reduction, while also minimizing long-term damage to the brain. The goal of this project is to automatically optimize a closed-loop strategy for the control of deep brain stimulation using reinforcement learning methods.  Reinforcement learning is an ongoing area of research in computer science, which can be used to learn and adapt stimulation strategies directly from electrophysiological recordings, and at various time-scales (from initiating a quick response to pre-seizure brain activity, to adapting to long-term changes in the epileptic disorder). We will test this new technology using both in vitro and in vivo models of limbic seizures.  In the longer term, we predict that this technology will play a key role in treating people with epilepsy, in particular patients presenting with mesial temporal lobe. In addition, we are confident that reinforcement learning could be integrated in a wide range of implantable medical stimulators, such as those currently used to treat Parkinson's, heart disease and  psychotic depression.